Targeted gene flow (TGF) could bolster the adaptive potential of isolated populations threatened by climate change, but could also lead to outbreeding depression. Here, we explore these possibilities by creating mixed- and within-population crosses in a terrestrial-breeding frog species threatened by a drying climate. We reared embryos on wet and dry soils and quantified fitness-related traits upon hatching. TGF produced mixed outcomes in hybrids which depended on crossing direction (origin of gametes from each sex). North-south crosses led to low embryonic survival if eggs were of a southern origin, and high malformation rates when eggs were from a northern population. Conversely, east-west crosses led to one instance of hybrid vigour, evident by increased fitness and desiccation tolerance of hybrid offspring relative to offspring produced from within-population crosses. These contrasting results highlight the need to experimentally evaluate the outcomes of TGF for focal species across generations prior to implementing management actions.

We explored the potential of targeted gene flow (TGF) to mitigate declines in population-level fitness using the crawling frog, Pseudophryne guentheri, a species in which populations are threatened by habitat loss and declining winter rainfall. We evaluated TGF within a laboratory setting by creating pure and reciprocal crosses among four geographically distant populations. Two populations occurred in low-rainfall regions at the northern edge of the species’ range, and two other populations were from higher-rainfall regions close to the centre the species range. We then assessed phenotypic traits in the resulting offspring, comparing individuals reared on wet soils (-10 kPa, a benign treatment) to those reared on drier soils (-400 kPa) that significantly reduce survival and hatchling fitness.

Detailed methods from manuscript:

We collected adult P. guentheri from four geographically separated breeding sites, situated at two latitudes, in May and June 2017 (Table 1). Sites spanned a ~460 mm annual rainfall gradient, with site A receiving the most rain per year and site D receiving the least (Table 1, Fig. 1). Pseudophryne guentheri collected from breeding populations at each site show variation in desiccation tolerance, with adults and embryos from site A being the most sensitive to dry conditions³⁷. Population genetic analysis³⁹ has demonstrated high levels of inbreeding in all populations (Table 1), and genetic differentiation among P. guentheri populations is high (overall F_ST = 0.186), which suggests low levels of contemporary dispersal. P. guentheri from sites A and B form distinct genetic clusters³⁹, indicating low historical gene flow despite their close proximity (100 km), whereas P. guentheri from sites C and D show admixture, but are genetically distinct from populations A and B³⁹.

In total, 15-16 calling males from each population were collected by hand and in pit-fall traps. Gravid females were more difficult to collect due to their cryptic behaviours, and so sampling was restricted to 5-13 females from each of three sites (A, B and C; Table 1). All frogs were temporarily housed in small (4.4 L) plastic terraria containing moist sphagnum moss, and transported to the University of Western Australia within two days of collection. There, frogs were fed a diet of pinhead crickets and kept in a controlled-temperature room at 16 °C with an 11/13 h light/dark photoperiod to mimic winter conditions.

Breeding design and in vitro fertilisations. Egg clutches of each female were divided equally into four groups, and fertilised with sperm from males originating from each of the four populations, resulting in one pure and three hybrid crosses. To control for potential parental compatibility (i.e. specific pairwise male-by-female) effects on offspring fitness^38,69, a sperm mixture, containing sperm from five random males from the appropriate population, was used to fertilise the eggs of each female in each population⁸⁸.

Sperm was obtained from testes macerates after euthanizing males via ventral immersion in <0.03% benzocaine solution, followed by double pithing. Sperm was stored on ice in 25-458 μL (adjusted according to the weight of the testes) standard amphibian ringer (SAR; 113mM NaCl, 2mM KCl, 1.35 mM CaCl₂). This buffer allows storage of sperm for extended periods (days–weeks) without substantial declines in motility^89,90. Sperm concentrations were measured using an improved Neubauer haemocytometer (Hirschmann Laborgeräte, Eberstadt, Germany) and sperm suspensions were diluted with 1:1 SAR to 100 sperm per μL.

Upon arrival at the laboratory, females were gently squeezed to determine whether ovulation had occurred. Approximately 35% of females had ovulated naturally while in transit and their eggs were gently stripped. For the remaining females, ovulation was induced via two subcutaneous injections of the hormone LHRHa over the course of two days^37,87. Approximately 10 hours after the second injection, eggs were gently stripped from each female. In all instances, freshly stripped eggs were moistened with SAR and distributed equally among four small petri dishes. A standardised number of sperm from five random males collected at each site was pipetted onto one edge of each petri dish, mixed gently with the pipette tip, and then activated with a pre-calculated volume of 1:4 SAR solution³⁸. This resulted in eggs from all females in the experiment being fertilised from males collected from sites A, B, C and D. Each dish was then manually agitated for 20 seconds to promote fertilisation. After 15 minutes, eggs were temporarily submerged in water, backlit and photographed using a digital imaging camera (Leica DFC320) attached to a light microscope (Leica MZ7.5) at 6.3 Χ magnification. These images were used to measure the ovum diameter of 50 randomly-selected eggs from each female, using ImageJ software⁹¹. Fertilisation success was initially scored one hour after mixing eggs and sperm by counting eggs that had rotated (Gosner⁷⁹ stage 1). However, eggs from populations A and B took substantially longer to show signs of fertilisation when mixed with sperm from populations C and D. We therefore scored fertilisation success a second time, six hours after sperm and eggs were mixed.

Incubation treatments. Fertilised eggs from each cross were reared on sandy loam soil at two water potentials (ψ): a wet soil (ψ = -10 kPa) and dry soil (ψ = -400 kPa). The soil was previously collected from a separate P. guentheri breeding site, and the soil water potentials represented a range found in natural nest sites (N. J. Mitchell, unpubl. data). Embryo incubation and soil preparation were performed as described in Rudin-Bitterli et al.⁹². Briefly, soil was oven-dried at 80°C for 24 hours, distributed into small containers and rewetted with an appropriate mass of deionised water using a water-retention curve previously determined for the soil sample (N. J. Mitchell, unpubl. data). The water content of the soil (g/g/ of oven dry soil) was approximately 50% in the wet treatment, and 21% in the dry treatment, and containers were sealed with a lid after wetting. Fertilised eggs from each cross were selected at random and distributed onto soils within 7 – 9 hours of fertilisation. Small plastic rings (nylon plumbing olives, 12 mm in diameter) were labelled and placed around eggs to identify individual crosses. Sealed containers were then placed in incubators set at 16 ± 0.5 °C, and embryos were monitored every two days. Any dead eggs were removed and discarded.

Response variables. Putative fitness from within- and between-population crosses, reared in dry and wet rearing environments, was assessed at hatching. At 33 days after fertilisation (when embryos were approximately at Gosner⁷⁹ Stage 26), hatching was induced by placing embryos individually in small test tubes containing 2 ml of deionised water³⁸. Embryos were then monitored at least every 30 min until hatching, defined as when an individual completely escaped their egg capsule. Embryonic survival was recorded for each family as the percentage of fertilised eggs that hatched.

Swimming performance was recorded 6 - 12 hours after hatching on a subset of hatchlings (N = 633 across all within and between-population crosses). For this purpose, individual hatchlings were placed in a petri dish (diameter = 150 mm) containing water 10 mm deep. After an initial acclimation period of 1 min, the tail of each hatchling was nudged with a glass cannula to elicit a burst swimming response. A video camera (Canon PowerShot G16, recording at 60 fps) installed 300 mm above the petri dish was used to film three burst swimming responses for each hatchling, and their movement was later tracked and analysed using EthoVision v8.5 software⁹³. EthoVision enabled the quantification of the following swimming parameters: maximum velocity (cm s^-1), mean velocity (cm s^-1) and total distance moved (cm). We also recorded mean meander (deg cm^-1), a measure of the straightness of the swimming response, as dry rearing environments can lead to asymmetrically shaped hatchlings^37,38 that swim in a more circular motion. A hatchling was considered to be moving when it exceeded 0.45 cm s^-1. As each video recording contained three burst swimming responses with periods of no movement in between them, EthoVision only analysed frames in which a hatchling moved faster than 0.45 cm s^-1 (consequently merging the three swimming responses for each hatchling). Immediately following the swimming performance trials, hatchlings were euthanized in <0.03% benzocaine and preserved in 10% neutral buffered formalin.

Wet masses of preserved hatchlings were recorded to the nearest 0.001 g after blotting on tissue. Hatchlings were then photographed in lateral view (while submerged in water to minimize refraction) using a digital imaging camera (Leica DFC320) attached to a light microscope (Leica MZ7.5) at X 6.3 magnification. These images were used to score malformations for each hatchling and to determine their developmental stage⁷⁹ by examining the hind limb buds.

Statistical analysis. All analyses were performed in R version 3.4.3 (R Development Core Team 2017). Linear mixed-effects models (with restricted maximum-likelihood methods; REML) were run using the lme4 package⁹⁴ to compare offspring traits from pure and hybrid crosses. In these models, treatment, female (population) origin, male (population) origin and all interactions (female origin-by-male origin, female origin-by-treatment, male origin-by-treatment, female origin-by-male origin-by-treatment) were considered as fixed factors. We checked for overdispersion in our models using the ‘overdisp_fun’ function proposed by Bolker et al ⁹⁵. Only one trait, embryonic survival, was overdispersed; we added observation level as an extra random factor to account for overdispersion when analyzing this trait (Harrison 2014). Because the eggs from each female were subjected to a split-clutch design, the term for dam (i.e. individual female ID) was added as a random effect to account for the use of individual females across multiple fertilisation events. We also included ovum size as a covariate in all analyses to control for possible maternal effects arising from different patterns of egg provisioning among females³⁸. The significance of the fixed effects was evaluated using Wald chi-squared tests.

Embryonic survival and hatchling malformation data were binomial variables and thus a generalized-linear mixed-effects model (GLMM) with a logit-link function was used for the analysis of these traits. In these models we included treatment, female (population) origin, male (population) origin and all interactions (see above) as fixed effects. As above, dam ID was treated as a random effect and ovum size was added as a covariate to control for possible maternal effects³⁸. The significance of the fixed effects was evaluated using Wald Z tests.